Climatic thresholds and temporal decoupling drive reproductive phenology in Neotropical cloud forests: a 36‐year analysis

Abstract

Reproductive phenology influences ecosystem functioning and climate sensitivity in tropical montane cloud forests, but phase‑specific climatic thresholds and the temporal coupling between flowering and fruiting remain poorly quantified over decadal scales. We analyzed a 36‑year monthly record (1983–2018) from a Venezuelan cloud forest (15 focal species) and NASA/POWER hydroclimatic data, integrating random forest models, regression‑tree threshold analysis, climate‑window sensitivity analyses, scale‑dependent coherence, species‑level trend analyses, and temporal models that account for seasonality and autocorrelation. In the complete seven‑variable model for 2001–2018, precipitation was the dominant concurrent predictor of both flowering and fruiting, with regression‑tree thresholds of 32.57 and 88.01 mm month −1 , respectively. A reduced model using only the three highest‑ranked predictors per phase recovered identical thresholds to the complete model; neither specification, however, remained stable under temporally structured cross‑validation. Sensitivity analyses supported the flowering threshold across model specifications, whereas fruiting was more model‑dependent and showed sensitivity to antecedent atmospheric conditions in the full period (excluding surface UVA irradiance). Flowering and fruiting showed strongest coherence at sub‑annual scales (peak at eight months), while long‑term trajectories diverged: flowering declined and fruiting occurrence increased; both trends remained significant under seasonality‑adjusted models with autocorrelation‑robust errors and AR(1) sensitivity analyses. Species‑level analyses showed that flowering decline was broadly distributed but not ubiquitous, whereas the increase in fruiting occurrence concentrated in a subset of taxa. These results indicate temporal decoupling: within‑year reproductive coordination is preserved, while long‑term reproductive occurrence is reorganized across species, predictor sets, and climatic windows.

The authors' abstract, as published at the source. Ecography, 2026 · DOI ↗